Evaluating the Substrate Selectivity of Alkyladenine DNA Glycosylase

Jan 14, 2016 - Dedon , P. C. and Tannenbaum , S. R. (2004) Reactive Nitrogen Species in the Chemical Biology of Inflammation Arch. Biochem. ...... Is ...
0 downloads 6 Views 3MB Size
Download PDF
Article pubs.acs.org/biochemistry

Evaluating the Substrate Selectivity of Alkyladenine DNA Glycosylase: The Synergistic Interplay of Active Site Flexibility and Water Reorganization Stefan A. P. Lenz and Stacey D. Wetmore* Department of Chemistry and Biochemistry, University of Lethbridge, 4401 University Drive West, Lethbridge, Alberta, Canada T1K 3M4 S Supporting Information *

ABSTRACT: Human alkyladenine DNA glycosylase (AAG) functions as part of the base excision repair (BER) pathway by cleaving the N-glycosidic bond that connects nucleobases to the sugar− phosphate backbone in DNA. AAG targets a range of structurally diverse purine lesions using nonspecific DNA−protein π−π interactions. Nevertheless, the enzyme discriminates against the natural purines and is inhibited by pyrimidine lesions. This study uses molecular dynamics simulations and seven different neutral or charged substrates, inhibitors, or canonical purines to probe how the bound nucleotide affects the conformation of the AAG active site, and the role of active site residues in dictating substrate selectivity. The neutral substrates form a common DNA−protein hydrogen bond, which results in a consistent active site conformation that maximizes π−π interactions between the aromatic residues and the nucleobase required for catalysis. Nevertheless, subtle differences in DNA−enzyme contacts for different neutral substrates explain observed differential catalytic efficiencies. In contrast, the exocyclic amino groups of the natural purines clash with active site residues, which leads to catalytically incompetent DNA−enzyme complexes due to significant reorganization of active site water. Specifically, water resides between the A nucleobase and the active site aromatic amino acids required for catalysis, while a shift in the position of the general base (E125) repositions (potentially nucleophilic) water away from G. Despite sharing common amino groups, the methyl substituents in cationic purine lesions (3MeA and 7MeG) exhibit repulsion with active site residues, which repositions the damaged bases in the active site in a manner that promotes their excision. Overall, we provide a structural explanation for the diverse yet discriminatory substrate selectivity of AAG and rationalize key kinetic data available for the enzyme. Specifically, our results highlight the complex interplay of many different DNA−protein interactions used by AAG to facilitate BER, as well as the crucial role of the general base and water (nucleophile) positioning. The insights gained from our work will aid the understanding of the function of other enzymes that use flexible active sites to exhibit diverse substrate specificity.

T

concentration causes A:T to G:C transition mutations,11 which are likely related to the observed inflammation and carcinogenesis associated with intestinal, liver, and brain tissues.12,13 Alternatively, the additional steric bulk at N3 of A halts standard DNA replication, and mutations are introduced when 3MeA is processed by the notoriously error prone translesion synthesis pathway,10,14 while 7MeG depurinates 106 times faster than G and leads to an increased presence of abasic sites within the genome.15 The base excision repair (BER) pathway removes lesions to maintain the fidelity of DNA.16−21 The DNA glycosylases initiate the BER pathway by scanning the genome to locate a lesion, flipping the damaged base into the active site, and catalyzing cleavage of the N-glycosidic bond connecting the nucleobase to the sugar−phosphate backbone, which generates

he DNA nucleobases are susceptible to chemical modification by a number of different intracellular and extracellular sources, which can lead to mutagenic or cytotoxic lesions.1,2 Indeed, endogenous reactive oxygen (i.e., hydrogen peroxide or hydroxyl radicals) or nitrogen (i.e., nitric oxide radicals) species can directly react with the DNA nucleobases to form common oxidation lesions.1−6 For example, adenine reacts with nitric oxide radicals to form hypoxanthine [Hx (Figure 1)].7 Alternatively, the DNA nucleobases can be alkylated through a variety of routes. For example, 1,N6ethenoadenine [εA (Figure 1)] and 3,N4-ethenocytosine [εC (Figure 1)] are formed from reactions between DNA nucleobases and products of lipid peroxidation.8,9 Alternatively, N3-methyladenine [3MeA (Figure 1)] arises when A reacts with endogenous S-adenosylmethionine, while N7-methylguanine [7MeG (Figure 1)] is formed when DNA is exposed to nitrosamines from tobacco smoke.10 If left unrepaired, DNA lesions can directly or indirectly cause mutations. For example, replication of DNA in mammalian cells with an increased εA © XXXX American Chemical Society

Received: October 29, 2015 Revised: January 5, 2016

A

DOI: 10.1021/acs.biochem.5b01179 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

Because of the observed AAG substrate diversity, it is unsurprising that X-ray crystal structures reveal a lack of direct contacts between the nucleobase of the bound nucleotide and active site amino acids (Figure 2).31,34,35 Instead, it has been

Figure 1. Structures and chemical numbering of the (A) canonical purines, (B) neutral purine lesions, (C) cationic purine lesions, and (D) the 2′-deoxyribose moiety. Nucleobase modifications are colored red.

an abasic site. In humans, alkyladenine DNA glycosylase (AAG) is responsible for the removal of a variety of purine lesions, including Hx, εA, 3MeA, and 7MeG.22−24 This substrate diversity is unusual among the glycosylase family, with most enzymes targeting a specific lesion.25−27 Indeed, it is difficult to pinpoint a common structural motif among AAG substrates, which vary in terms of the charge, size, and position of hydrogen bond donor/acceptor atoms. As a consequence, AAG sacrifices catalytic efficiency,22,23 exhibiting modest catalytic rate enhancements (103−107) compared to those of other glycosylases (1010−1026).23,24,28,29 Despite the structural diversity among AAG substrates, the enzyme does not catalyze the removal of pyrimidine lesions, with εC inhibiting enzyme activity.30,31 Furthermore, although AAG has been shown to bind mismatched A or G, the enzyme exhibits very low catalytic rates when targeting the canonical purines.23 The specificity of DNA glycosylases, including AAG, is at least in part dictated by the base-flipping step during which the enzyme inserts an amino acid residue into the DNA helix to facilitate flipping of the damaged nucleotide into the active site.16−21 For example, because uracil DNA glycosylase (UDG) has a narrow binding pocket that excludes purine substrates, the base-flipping step significantly contributes to the high specificity of this repair enzyme for uracil.32 AAG activity has also been confirmed to be affected by the base-flipping step.23 For example, Hx is excised at a higher rate when paired opposite T than opposite C,23 which has been attributed to the lower stability and therefore ease of base flipping in the Hx:T base pair than in the Hx:C base pair.33 Similarly, AAG exhibits activity toward the natural purines in destabilized mismatched base pairs greater than that of canonical base pairs.23 Nevertheless, AAG exhibits a catalytic rate several orders of magnitude higher for the excision of damaged nucleobases than for mismatched natural purines.23 Therefore, although base flipping is one important factor, the substrate specificity of AAG is also partially determined by the chemical step. Indeed, the excision of εA is minimally affected by the identity of the opposing base,23 and the associated rate-limiting step has been determined to be glycosidic bond cleavage.24

Figure 2. (A) X-ray crystal structure of AAG bound to εA (PDB entry 1EWN).35 (B) Overlay of the AAG active site bound to εA (gray, PDB entry 1EWN)35 or εC (orange, PDB entry 3QI5).31 Key hydrogen bonds with the nucleobase are highlighted with dashed lines.

proposed that AAG primarily uses nonspecific π−π interactions [involving Y127, H136, and Y159 (Figure 2A)] to position damaged nucleobases in the active site. A previous computational study suggests that these active site π−π interactions play an additional role, being catalytic toward the removal of neutral nucleobases (εA) but noncatalytic or even anticatalytic toward excision of cationic lesions (3MeA).36 This multifold function of π−π interactions may not be surprising given that DNA− protein π−π contacts have been shown to be prevalent across biological systems and provide stability comparable to that of specific hydrogen bonding interactions.37,38 Another critical AAG active site residue is E125 (Figure 2A), which has been proposed to activate the water nucleophile that cleaves the glycosidic bond in the damaged nucleotide via a substitution reaction.34 Indeed, computational work characterized an SN2 hydrolysis mechanism facilitated by E125 for the removal of both neutral and cationic AAG substrates.36 The proximity of E125 to the site of reaction (C1′) is also critical to glycosylase activity because this negatively charged residue stabilizes the positive charge that accumulates on the deoxyribose moiety during deglycosylation. Indeed, E125Q and E125A mutations abolish AAG activity.22 Finally, although it has been proposed that AAG requires a general acid near N7 of the nucleobase for catalysis via leaving group activation, no active site amino acid capable of this function has been identified.22 Furthermore, despite proposals that a water chain may play the role of the general acid,35 combined mass spectrometry and computational studies indicate that (full) nucleobase protonation may not be a requirement for base excision.39,40 Although important mechanistic information has been acquired to date on AAG, atomic level structural information that explains the broad AAG substrate specificity, and the ability B

DOI: 10.1021/acs.biochem.5b01179 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

To explain the diverse yet discriminatory substrate selectivity of AAG, this study employs molecular dynamics simulations to evaluate how changes in the chemical structure of bound nucleotides affect the AAG active site conformation and key active site interactions. Specifically, seven nucleotides were examined (Figure 1) that are relevant to AAG, including neutral substrates (εA and Hx), cationic substrates (3MeA and 7MeG), an inhibitor (εC), and the canonical purines (A and G). By uniquely considering a large range of nucleotides and monitoring active site reorganization following DNA−enzyme complex formation, we provide the first structural explanations for how AAG can achieve broad substrate diversity and simultaneously avoid excising the natural purines. Furthermore, each class of nucleotides considered yields important clues regarding the roles played by active site residues, as well as water, and reveals that these roles can significantly vary depending on the substrate bound. In fact, subtle deviations in active site interactions provide insight into experimentally observed differences in catalytic efficiencies toward excision of various substrates. In addition to affording a more thorough understanding of this critical DNA repair enzyme, our study contributes more broadly to the role that active site reorganization and placement of active site water play in substrate or inhibitor recognition and binding, and therefore is complementary to previous work on a wide variety of other enzymes, including anthipeptide synthetases,59 metalloenzymes,60 aminoacyl tRNA synthetases,61 and both DNA and RNA polymerases.62,63

of the enzyme to discriminate against the canonical purines, is still missing. Close comparison of crystal structures of AAG bound to the εA substrate and εC inhibitor (Figure 2B) reveals key differences in binding that (at least partially) rationalize the differential AAG activity toward these lesions.31,35 Specifically, although both nucleobases form a hydrogen bond with the backbone of H136, εC is pulled farther into the active site of the enzyme, and an additional hydrogen bond is formed between εC and the side chain of N169 (Figure 2B). The extra hydrogen bond in the case of εC has been proposed to explain the observed binding affinity of AAG for εC that is 2-fold greater than that for εA (that is lost upon N169L or N169I mutation), while enzyme inhibition was proposed to be caused by the lack of leaving group activation.31 Interestingly, there are few overall changes to the configuration of AAG active site amino acids upon binding of εA and εC. However, both nucleotides interact with H136 through a similarly positioned hydrogen bond acceptor [N2 for εC and N6 for εA (Figure 2 and Figure S1)], and not all nucleotides can form this interaction. Indeed, for A, it has been proposed that steric clashes between the N6 amino group and the H136 side chain prevent AAG activity at an appreciable rate.23 Nevertheless, 3MeA possesses the same N6 amino group as A yet is excised by AAG.23 This suggests that substrates may bind to the active site in different ways, and active site reorganization may be important for understanding the relative activity that AAG has toward different nucleotides. Indeed, recent experimental evidence indicates that at least part of the AAG active site is disordered prior to substrate binding and suggests that the final active site conformation may be dictated by the nucleotide bound.41 However, it is difficult to predetermine the changes that may occur when AAG interacts with chemically diverse nucleotides. Despite the importance of understanding how different nucleotides bind to AAG, only DNA containing εA or εC has been successfully cocrystallized with AAG to yield structures with intact glycosidic bonds. Indeed, experimental challenges arise because, for example, cationic lesions have short halflives,16 while AAG does not bind the canonical purines at an appreciable rate to allow crystallization.23 However, computational chemistry has proven to be a useful approach for obtaining molecular level structural information about biochemical systems (see, for example, refs 42−49), including DNA glycosylases.36,50−57 Indeed, computational studies have considered certain aspects of the AAG mechanism of action for select nucleotides.36,50,58 Specifically, large-scale ONIOM models were used to gain mechanistic information about AAG activity toward εA, 3MeA, and A,36 while a range of Monte Carlo (MC)50 and molecular dynamics (MD)58 simulation methods were used to consider the binding modes of A, Hx, εA, and 1,N6-ethanoadenine (EA). Although these studies provided useful information about the catalytic mechanism or key interactions with AAG for specific substrates, information about differences in the active site conformation upon binding was not explicitly obtained for a range of nucleotides that span the chemical diversity of AAG substrates due to the computational models employed [i.e., ONIOM models based on a single (static) crystal structure or constraints applied to the DNA and protein backbones] or the structural similarity of the substrates considered (i.e., εA and EA can both interact with H136). Therefore, more work is required to understand the function of AAG.

■

COMPUTATIONAL DETAILS Two X-ray crystal structures of AAG available in the PDB were used to initiate MD simulations, namely, Δ79AAG, which contains an active site E125Q mutation bound to the εA substrate (PDB entry 1EWN),35 and Δ80AAG, which contains the enzyme bound to the εC inhibitor (PDB entry 3QI5).31,35 Both structures were chosen because of their high resolution (2.1−2.2 Å), and the fact that the enzyme is bound to a lesion with an intact glycosidic bond (i.e., can be used to probe substrate binding). Both Δ79 and Δ80AAG (henceforth termed AAG) were truncated for ease of crystallization but have catalytic power equivalent to that of AAG.34 Although 1EWN contains a single AAG E125Q−εA complex in the unit cell, 3QI5 includes two AAG−εC complexes, with the complex containing the fewest unresolved residues used for our study. 1EWN has unresolved density for residues 80, 81, 200−207, 249−254, and 295−298, while 3QI5 has unresolved density for residues 205, 206, 265, 266, and 294−298. Overlays of the two complexes were used to assign the position and orientation of residues resolved in only one of the structures, while the backbone was added by inspection when a residue was unresolved in both structures. This approach is justified because all unresolved residues are remote from the active site (>10 Å separation). The LEAP module of AMBER 10 was used to add the unresolved side chains and hydrogen atoms to all residues.64 The locations of the added side chains were manually inspected using PyMol and altered to favor hydrogen bonding interactions and minimize steric clashes. The E125Q mutation in 1EWN was reverted to generate the wild-type enzyme. To consider the impact of different nucleotides, other DNA−protein complexes were generated by replacing εA in the structure generated from 1EWN with Hx, A, G, 3MeA, or 7MeG. C

DOI: 10.1021/acs.biochem.5b01179 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

εA or εC bound to AAG yield active site conformations consistent with the crystallographic data (Figure S1). Specifically, the average rmsd across the simulation trajectory relative to the corresponding crystal structure is less than approximately 0.8 Å (Table S1). The average rmsd is largest for εA due to reversion of the experimental E125Q mutation in the MD simulation, which yields an E125 position consistent with that observed in the crystal structure of the wild-type enzyme bound to εC (Figure S1). Upon comparison of the simulated structures with AAG bound to εA, εC, or Hx, the active site conformation does not change significantly regardless of the nucleotide bound (Figure 3A), with the average rmsd relative

Each DNA−protein system was assigned AMBER parm99SB parameters65 supplemented with GAFF parameters66,67 for nonstandard nucleotides. Additionally, restricted electrostatic potential (RESP) charges were assigned to each nonstandard nucleoside using the R.E.D.v.III.4 program68 and B3LYP/631G(d)-optimized geometries obtained with Gaussian 09 (revision C.01 or D.01).69 Each DNA−protein system was neutralized with Na+ counterions and solvated in an explicit TIP3P water box, which ensured that the DNA−protein solute was no less than 8.0 Å from the boundary of the water box in any direction,64,67 and the particle mesh Ewald method was used with an electrostatic cutoff of 12.0 Å. Using a 2 fs time step, the solvent and ions were minimized for 500 steps of steepest descent and 500 steps of conjugate gradient minimization, while applying a 500.0 kcal mol−1 Å−2 constraint to the protein and DNA. Subsequently, the system was minimized using 1000 steps of steepest descent, followed by 1500 steps of conjugate gradient optimization. The system was then heated to 300 K over 20 ps with a 10 kcal mol−1 Å−2 restraint placed on the solute. Finally, the entire system was simulated without constraints under NPT conditions (1 atm, 300 K) for 40 ns. All minimization and simulation steps were performed with SANDER and PMEMD,64,70 and subsequent analysis was completed using the cpptraj module. Free energy calculations were performed using the molecular mechanics generalized Born surface area (MM-GBSA) method available in Amber Tools.64 The reported pairwise energies were calculated between each active site amino acid and the bound nucleotide. To monitor changes in the active site structure, the average root-mean-square deviation (rmsd) and associated standard deviation (σ) in the active site across two simulation trajectories or a single trajectory relative to a static crystal structure were calculated using the position of the heavy atoms in both the side chains and backbones of the bound nucleotide (dX), and active site (E125, Y127, A135, H136, Y159, N169, L180, and R182). Specifically, root-mean-square (rms) fitting was performed on the basis of the position of these residues to yield the overall rmsd of the active site, and subsequently, the rmsd per residue was determined without refitting. All reported distances were measured and averaged throughout each simulation trajectory. The reported intermolecular separations between the nucleobase and amino acid rings correspond to the distances between the centers of mass of the aromatic moieties. Structures reported in the main text were obtained by clustering the position of the bound nucleotide and the active site residues and choosing a representative structure from the cluster with the greatest occupancy. Although the reported representative structures yield static pictures, we report the corresponding dynamic data across each trajectory in the Supporting Information. To determine the distribution of water in the active site, a three-dimensional 20 Å × 20 Å × 20 Å grid was centered on the active site with 0.50 Å spacing between each point on the grid using Amber Tools.62 Spheres of different colors are shown on representative structures to reflect the percentage of each trajectory that an oxygen atom of a water molecule occupied a 0.50 Å × 0.50 Å × 0.50 Å grid space, with gray spheres representing 40% occupancy, yellow spheres 60% occupancy, and red spheres 80% occupancy.

Figure 3. Overlay of representative MD structures of AAG bound to εA (gray), εC (orange), or Hx (blue) highlighting (A) the consistent active site conformation for the neutral lesions and (B) the common hydrogen bond between the H136 backbone and the bound nucleobase (dX).

to bound εA being 0.981 Å for Hx and 1.433 Å for εC (Table S2). The average rmsd highlights the similarity in the binding of the neutral substrates εA and Hx, while the rmsd is larger in the case of the εC inhibitor because the lesion is further inserted into the active site as previously discussed on the basis of experimental crystal structures.31,35 Nevertheless, the relative orientations of the aromatic (Y127, H136, and Y159) residues and the bound lesions do not change substantially [key torsional angles are within 30° (Table S3 and Figure 3)]. Furthermore, the relative orientation of the E125 general base with respect to the bound lesion is consistent [within 7° (Table S3 and Figure 3A)], and a hydrogen bond is maintained in the E125-Y127 catalytic dyad [>99% occupancy (Table S4)] regardless of the neutral lesion considered. However, the average distance between E125 and C1′ deviates by up to 0.830 Å (Table S5), being the longest for εA (4.758 Å) and shortest for Hx (3.928 Å). Furthermore, the placement of water within the active site deviates when the neutral lesions are bound to AAG (Figure S2A−C), with a distinct lack of water between the substrate and the E125 general base for εC. In contrast, there is substantial water density near E125 [>100% occupancy of the E125(Oε)···H2O(OH) hydrogen bond (Table S4)], which

■

RESULTS AAG Adopts a Consistent Active Site Conformation upon Binding a Range of Neutral Lesions. MD simulations initiated from X-ray crystal structures with either D

DOI: 10.1021/acs.biochem.5b01179 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry could correspond to the nucleophile, when either εA or Hx is bound to AAG. The absence of nucleophilic water may be one factor in εC inhibition of AAG activity. Despite AAG lacking many direct contacts with the bound lesion, it has been noted on the basis of experimental crystal structures that both εC and εA have a hydrogen bond acceptor [N4 for εC and N6 for εA (Figure 1)] well positioned to interact with the NH backbone of H136 (Figure 2B).31,35 Overlays of representative MD structures obtained with εA, εC, or Hx bound to AAG reveal that O6 of Hx adopts an equivalent acceptor position (Figure 3B). Indeed, the dX(N2/N4/O6)··· H136(NH) hydrogen bond is occupied for >98% of the simulation trajectory corresponding to each neutral lesion (Table S4) and, in combination with a dX(O1P)···H136(NεH) interaction, contributes between −7.0 and −10.0 kcal mol−1 to the binding free energy between the nucleotide and AAG (Table S6). Additionally, the contact with H136 leads to the observed further insertion of εC into the active site because of the shorter distance between the hydrogen bond acceptor (N2/ N4/O6) and the glycosidic bond (N1/N9) in εC compared to εA and Hx (Figure 3A). Other interactions also occur for at least 80% of the simulations, namely, the dX(O1P)··· H136(NεH) hydrogen bond when εA or εC is bound and the dX(O3)···R182(NηH) hydrogen bond when Hx is bound (Table S4), which further facilitates nucleotide binding by contributing up to −12 kcal mol−1 to the binding free energy (Table S6). However, when εC or Hx is bound to AAG, an additional hydrogen bond forms between the damaged nucleobase and an active site residue (Table S4). Specifically, N169 (NεH) hydrogen bonds with O2 of εC (44.8% occupancy and binding energy of −2.73 kcal mol−1), and Y159 (OH) hydrogen bonds with N3 of Hx [73.1% occupancy (Table S4) and binding energy of −2.57 kcal mol−1 (Table S6 and Figure S3)]. Despite the formation of a unique hydrogen bond with εC and Hx, the active site conformation adopted by AAG is nearly equivalent when it is bound to any of these neutral lesions. Because the common hydrogen bond contact between each damaged nucleobase and H136 likely plays a significant role in stabilizing the consistent active site configuration, it is worthwhile to examine the structural consequences when nucleotides that lack the ability to hydrogen bond with the H136 backbone bind to AAG. Reorganization of the AAG Active Site, Including Increased Solvation, Occurs upon Binding of Adenine and Likely Prevents Excision. When A is bound to AAG, the nucleotide cannot form a hydrogen bond with the H136 backbone, because the N6 exocyclic amino group occupies the position of the hydrogen bond acceptor in the neutral lesions (Figure 1). Because the neutral lesions induce a similar AAG active site conformation, we focus our comparison on the binding of A and Hx because of their similar size, and a key difference of the C6 substituent. Overlays of representative MD structures with A and Hx bound to AAG (Figure 4A) reveal that the relative orientations of E125, Y127, and L180 with respect to the nucleotide are consistent (Figure 4A). Specifically, the average rmsd for each of the aforementioned residues is 99% occupancy) is absent when G is bound to AAG [99% occupancy (Table S4)]. Indeed, the binding free energy between G and H136 (−9.6 kcal mol−1) is similar to that for the neutral lesions [−7.8 to −10.5 kcal mol−1 (Table S6)]. Furthermore, there is no active site water between the G nucleobase and the aromatic residues (Figure S2E). The major differences in the conformation of the AAG active site when bound to G versus Hx arise in N169 and E125 (Figure 5A). Specifically, a significant rmsd for N169 [1.377 Å (Table S2)] arises due to a hydrogen bond between the N2 amino group of G and the carbonyl of N169 [>99% occupancy (Table S4 and Figure 5B)]. This interaction is likely

Figure 6. Overlays of representative MD structures of AAG bound to (A) 3MeA (green) and A (light blue) or (B) 7MeG (green) and G (orange).

including a hydrogen bond to Y127 [>99% occupancy (Table S4)]. However, the average distance between E125 and the (C1′) reaction center is greater for 3MeA [by on average 0.533 Å (Table S5)]. This occurs because the damaged nucleotide is less inserted into the AAG active site (Figure 6A) due to steric repulsion between the 3MeA methyl group and the L180 side chain. This altered nucleotide position ensures that the steric F

DOI: 10.1021/acs.biochem.5b01179 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

Despite Similar Active Site Configurations, Subtle Disparities in DNA−Protein Interactions Lead to Different Catalytic Efficiencies toward the Excision of Neutral Substrates. A crystal structure of εA bound to AAG (PDB entry 1EWN) highlights a close contact between N6 of the nucleobase and the NH backbone of H136.35 Furthermore, superposition of Hx onto εA led to speculation that Hx could also adopt a position in the AAG active site to form an analogous hydrogen bond between O6 of the nucleobase and H136,23 which was more recently confirmed with Monte Carlo simulations using a constrained DNA−enzyme model.50 Similarly, a hydrogen bond to H136 is maintained for both εA and Hx throughout our unconstrained MD simulations (Table S4). More importantly, our simulations reveal that this interaction results in a consistent orientation of the H136 side chain, as well as Y127 and Y159, relative to the nucleobase regardless of whether εA or Hx is bound to AAG (Tables S2 and S3 and Figure 3). The uniform relative positioning of the substrate and aromatic amino acids is likely critical for the excision of neutral substrates. Indeed, previous computational work has shown that π−π interactions facilitate removal of εA by AAG.36 Furthermore, the strength of DNA−protein π−π interactions is comparable to that of biologically relevant hydrogen bonds, yet highly dependent on the proximity and orientation of the DNA and protein moieties.36−38 Other known neutral AAG substrates, including 1,N2-ethenoguanine and EA,30 also have a hydrogen bond acceptor ideally positioned to form an interaction with the H136 backbone and therefore will likely result in a similar AAG active site conformation upon binding. Indeed, previous computational work illustrates that EA binds to AAG in a manner comparable to that of εA.58 Regardless of the analogous AAG active site conformation upon binding of εA and Hx, experimental studies have reported that AAG binds εA more tightly than Hx yet excises Hx at a higher rate.23 Our MD simulations provide a molecular explanation for this intriguing observation in two ways. First, a hydrogen bond forms between the exocyclic hydroxyl group of Y159 and N3 of Hx that is not observed for εA (Figure 3B). This interaction will facilitate Hx excision by stabilizing the negative charge accumulating on the nucleobase upon deglycosylation. Indeed, hydrogen bonding has been shown to be used by other DNA glycosylases in lieu of protonation to stabilize the departing nucleobase.18,26,32 In contrast, Y159 forms a hydrogen bond to the DNA phosphate backbone when εA is bound to AAG (Table S4). The stronger interactions with the negatively charged phosphate in the case of εA (Tables S4 and S6) contribute to the observed stronger binding of εA versus Hx but will not assist base excision. Second, E125 is closer to the reaction center (C1′) for Hx than for εA (Table S5). A conserved E or D residue has been shown to be important for base excision facilitated by other DNA glycosylases, including hUNG2, 51,71 MutY, 25,52 and hOgg1,54,72 with the proposed role being stabilization of the positive charge developing on the sugar during base departure. Indeed, E125Q and E125A AAG mutants lack catalytic activity.22 Therefore, the closer proximity of E125 to the reaction center when Hx is bound likely also contributes to the observed greater catalytic rate enhancement for this substrate. Overall, despite neutral lesions binding in a similar fashion to AAG, subtle differences in active site interactions due to varied nucleotide composition play a role in dictating relative binding strengths and catalytic efficiencies.

clash between the A amino group and the H136 backbone does not occur for 3MeA. As a result, the average rmsd between the AAG active site conformations with 3MeA or A bound is 2.592 Å (Table S2). The active site residue that most significantly contributes to the large rmsd is H136 [2.266 Å (Table S2)], with the relative orientation of the side chain (χ1) differing by approximately 100° (Table S3 and Figure 6A) and the aromatic amino acid being on average closer to 3MeA (5.224 Å) than A [6.325 Å (Table S5)]. Because of these differences in the active site conformation, water does not reside between 3MeA and the aromatic residues (Figure S2F). Therefore, the contribution of H136 to the binding free energy is greater for 3MeA (−4.3 kcal mol−1) than for A [−2.7 kcal mol−1 (Table S6)]. Thus, despite the presence of the N6 amino group, L180 favorably positions the cationic 3MeA lesion in the AAG active site for base excision. As discussed for 3MeA, the methyl group of 7MeG ensures that the lesion is less inserted into the AAG active site than G (Figure 6B). Specifically, a steric clash between the methyl substituent and the H136 side chain prevents 7MeG from adopting the position equivalent to that of G. Nevertheless, 7MeG and G both form favorable hydrogen bonds with the H136 backbone and N169 side chain [>99% occupancy (Table S4)] and binding free energies of up to −10.0 kcal mol−1 (Table S6). Furthermore, the aromatic active site residues (H136, Y127, and Y159) adopt consistent orientations relative to the nucleobase (Table S2 and Figure 6B), and key backbone torsional angles are within 20° for both nucleotides (Table S3). These similarities lead to a small rmsd in the AAG active site conformation when 7MeG or G is bound [0.936 Å (Table S2)]. However, a crucial difference in the AAG active site conformation is the maintenance of the hydrogen bond in the E125-Y127 catalytic dyad for 7MeG [>99% occupancy (Table S4)]. Furthermore, water is suitably located between E125 and C1′ for nucleophilic attack when 7MeG is bound to AAG (Figure S2G). Thus, despite the similar position of the N2 amino group in 7MeG and G, the methyl groups in the damaged nucleobases exhibit repulsion with an active site residue (H136), which thereby favorably positions the cationic lesions for catalytic excision by AAG.

■

DISCUSSION The major goal of this study was to examine the manner in which a number of different nucleotides can be bound within the active site of AAG and thereby uncover the structural basis for the ability of AAG to excise a structurally diverse set of substrates, while discriminating against the canonical purines. In this light, we examined seven different nucleotides, including two neutral substrates (εA and Hx), an inhibitor (εC), the canonical purines (A and G), and two cationic substrates (3MeA and 7MeG). Although previous literature suggests that the base-flipping step is one factor contributing to the substrate diversity of AAG,23,24 our model solely provides valuable structural information regarding how different nucleotides bind when forced into the active site. This information is critical because of the role of other factors in determining the observed AAG substrate diversity and differential catalytic rates, including the chemical step.23,24 Our results suggest that active site reorganization upon substrate binding allows the enzyme to exploit several active site residues, as well as water, to achieve broad yet discriminatory substrate selectivity and diverse catalytic efficiency. G

DOI: 10.1021/acs.biochem.5b01179 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry In Addition to a Key Active Site Hydrogen Bond, Redistribution of Water upon εC Binding Leads to Inhibition of AAG. Experimental studies have attributed the observed 2-fold stronger binding of εC versus that of εA to a hydrogen bond between N169 and O2 of εC.31,73 Although this is supported by the N169L and N169A mutants exhibiting decreased affinity for the εC lesion,73 the only structural support for this proposal came from a single static crystal structure of εC bound to the E125Q AAG mutant. Our MD simulations confirm that this hydrogen bond is significantly populated even when the dynamics of the active site are considered [45% occupancy (Table S4 and Figure S2A)], highlighting its importance for binding. Furthermore, the hydrogen bond between the nucleobase and the backbone of H136 and the relative orientations of the aromatic amino acids discussed for the neutral substrates are maintained when εC is bound to AAG (Figure 3), which further underscores the role of these residues in binding neutral nucleotides. The inhibitory behavior of εC has been attributed to the lack of nucleobase activation due to the absence of a proton acceptor equivalent to the N7 position of εA.31 However, experimental and computational studies indicate that protonation at the N7 site may not be required for glycosidic bond cleavage in εA36,39 or Hx,39 and neither A nor G is excised appreciably by AAG despite possessing an N7 acceptor site.23 Furthermore, although the mechanism for εC excision was not explicitly considered, previous computational work speculated that poor nucleophile activation could be a contributing factor to inhibition based on anticipated similarities to the mapped A excision mechanism.36 In contrast to these proposals, we attribute the inhibitory behavior of εC to the lack of water near the general base (E125) that could adopt the role of the nucleophile. Indeed, E125 interacts with water for